45 results about "Temporal consistency" patented technology

The present invention relates to an event-driven Extract-Transform-Load processing system and corresponding method that enable real-time synchronization of data originating from multiple heterogeneous and asynchronously operating data sources. The invention implements a machine-implemented processing approach in which data events dynamically trigger transformation determination, synchronization execution, and multi-stage computational validation without reliance on predefined batch schedules. Transformation operations are adaptively selected based on contextual data characteristics, historical synchronization behavior, and dynamically adjustable precision thresholds, thereby ensuring accurate data alignment under evolving data structures and operating conditions. The system further performs sequential validation of structural conformity, temporal consistency, and cross-source relational integrity prior to permitting data persistence or downstream transmission.

An electronic control unit in which a plurality of workloads are identified from execution of a runtime application, where each identified workload is associated with a set of cache lines of a cache resource. A priority designation of the set of cache lines is determined based at least in part on a usage of the set of cache lines during an upcoming time interval. In examples, the usage is based at least in part on memory access operations that are expected to be performed by the plurality of workloads during the upcoming time interval.

This invention relates to the field of water resource optimization scheduling technology in water conservancy and hydropower engineering, specifically to a reservoir optimization scheduling method and related products based on intertemporal sensitivity sparse repair. The method involves constructing and solving implicit equations for power generation and water volume, establishing an evaluation system and calculating intertemporal sensitivity to obtain a set of key time periods. Then, sparse repair is implemented on the initial power generation plan within this set, and the repair operation is embedded in an iterative optimization framework. This invention avoids the use of empirical estimation or coarse trial calculations in traditional reservoir scheduling, enhances the water resource safety net of reservoirs, suppresses the oscillation phenomenon of traditional intelligent evolutionary algorithms, and maintains the scheduling intent and temporal consistency of the initial predetermined plan on the power grid side.

This invention discloses a secure data transmission access system and method for intelligent medical IoT, relating to the field of data transmission control technology. It includes: S1, real-time acquisition of medical monitoring data and transmission information, and data preprocessing; S2, assessment of the temporal consistency of medical monitoring data sequences from each transmission channel, and data rearrangement and completion processing for transmission channels with sequence distortion; S3, analysis of the correlation of data change trends between different transmission channels, assessment of the time synchronization relationship between channels, and time shift alignment for channels with stable synchronization relationships; S4, construction of a patient physiological state representation, assessment of the coupling consistency between various physiological parameters, and determination of whether to allow subsequent patient state snapshot construction. This solves the problem of inaccurate physiological state representation caused by temporal disorder and cross-channel synchronization distortion during multi-source medical monitoring data transmission in an intelligent medical IoT environment.

The application discloses a remote control and data interaction method and system of a skid loader based on edge computing, and relates to the technical field of intelligent control of engineering machinery, and comprises the following steps: in view of the data interaction of multiple edge nodes in the remote control process of the skid loader, the sampling period and the instruction sending period of each edge node are sorted respectively, a unified time reference is established, and a time sequence is formed. Through the establishment of the unified time reference and the continuous time sequence, the time consistency between the multiple edge nodes is realized, the instruction flow fault is avoided, and the smooth transition of actions is ensured; meanwhile, through the construction of a cache misplacement comparison table and the re-planning of the data refresh sequence, the cache update is synchronized, the instruction transmission is continuous, and therefore the response continuity and the control stability of the remote control of the skid loader are improved.

This invention discloses a temporally enhanced knowledge graph-driven dialogue agent and its implementation method, belonging to the fields of artificial intelligence and natural language processing technology. The dialogue agent includes a dialogue interface, natural language understanding, a temporal knowledge graph, a temporal reasoning engine, dynamic update and failure management, and a natural language generation module. It includes designing the basic representation unit of the knowledge graph as a temporal quintuple; designing a temporal reasoning engine and a dynamic update and failure management method; designing a method for matching, verifying, and performing multi-hop reasoning of temporal information in natural language problems with temporal facts in the knowledge graph; achieving accurate parsing of temporal constraints in dialogue, real-time adaptation of dynamic knowledge, rigorous temporal logic reasoning, and continuous maintenance of contextual temporal consistency; and outputting a temporally accurate, knowledge-reliable, and logically coherent intelligent response, capable of meeting the needs of complex scenarios such as vertical industry decision support and high-reliability intelligent interaction.

The present application belongs to the field of computer vision and artificial intelligence, and particularly relates to a four-dimensional dynamic scene representation and reconstruction method, system and device based on structural equivalence prior, aiming at solving the problems of poor spatio-temporal consistency, large calculation cost and low long sequence reconstruction accuracy in dynamic scene reconstruction. The present application comprises: decomposing the input four-dimensional voxel sequence into multiple two-dimensional feature planes containing spatial projection and spatio-temporal coupling; introducing structural equivalence prior to fuse the plane features, generating unified and continuous spatio-temporal latent space representation by establishing the internal correlation between spatial geometric gradient and time evolution; finally, using a decoding module containing residual connection to decode the latent space representation, and recovering the high-fidelity four-dimensional dynamic scene layer by layer. The present application effectively improves the consistency and accuracy of dynamic scene reconstruction by establishing an explicit spatio-temporal coupling relationship, while taking into account the calculation efficiency, and has broad application prospects in the fields of autonomous driving, augmented reality, etc.

This invention provides a dynamic adaptive prediction method for bearing remaining life oriented towards dual-domain distribution offset, relating to the fields of fault prediction and health management. This invention is a remaining life prediction method capable of online sensing of data distribution changes and possessing safe adaptive capabilities. Specifically, it proposes an end-to-end algorithm implementation scheme. Its core idea is to construct a system with online sensing, decision-making, and updating capabilities to achieve continuous and safe optimization of the model after deployment. The overall technical solution is completed collaboratively by a time consistency comparison generalization training module, a multi-scale health state sensing module, and an uncertainty-guided safe adaptive update module. These three core modules are sequentially connected, forming a complete closed loop from offline pre-training, online state assessment to dynamic model updating.

This invention discloses a cross-market information propagation optimization method based on hexagram position entanglement, specifically relating to the fields of information processing and data analysis. By performing time alignment and semantic embedding on multi-source information streams, a unified basic information feature vector is constructed, and further, a hexagram position state vector including positional parameters, sentiment parameters, and propagation potential parameters is established. Based on this, by calculating the coherence and coupling strength between different information sources, a hexagram position entanglement relationship matrix is ​​constructed, and entanglement feature parameters are extracted. Propagation evolution is predicted by combining historical propagation paths, obtaining propagation distortion and micro-scale delay offset features, and thereby achieving adaptive adjustment of propagation weights and time windows. This method can improve the temporal consistency and prediction accuracy of cross-market information propagation and is suitable for propagation optimization analysis in complex multi-source information environments.

The present application relates to the field of machine vision detection, and particularly relates to a sealing size deviation real-time detection method based on machine vision, which comprises the following steps: firstly, obtaining an image in an uncompressed state and establishing an arc length coordinate; then, applying two-stage compression load according to a detection formula, collecting contact information, edge appearance information and displacement information to form a response sequence; subsequently, determining a local cross-section state sequence, and performing time consistency check and space consistency check; and finally, outputting a size deviation detection result and a process adjustment instruction. The present application can improve the accuracy of abnormal identification, the stability of position correspondence, and the timeliness of process correction.

This invention discloses a visual counting method and apparatus for combined scales. By establishing a mapping relationship between image coordinates and the spatial coordinates of the discharge port, it identifies occlusion conflict areas caused by simultaneous discharge from multiple channels. Differentiated acquisition frame rates are assigned to each channel based on the degree of conflict. The occluded segments in the material's movement trajectory are traced backwards to extract motion features and predict missing trajectory segments. The channel to which the material belongs is determined by combining the endpoint position and movement direction. Temporal cross-validation is performed on the count values ​​of each channel to identify counting deviations. Channels are classified into reliability levels based on the degree and trend of deviation, and multi-channel count values ​​are fused using a weighted classification system. Finally, dynamic correction parameters are generated through temporal consistency detection to correct abnormal jumps and output structured counting results. This invention achieves optimized allocation of sampling resources, accurate attribution determination under occlusion conditions, and reliable fusion of multi-channel data.

This invention discloses a method and system for intelligent construction of dynamic knowledge graphs based on a large model, relating to the field of artificial intelligence technology. The method includes: constructing a dynamic knowledge graph structure based on a dual-temporal data model to obtain a knowledge graph framework; extracting entities, relationships between entities, and corresponding temporal attribute information from input data using a large model; matching, merging, and deduplicating the extracted entities with existing entities in the dynamic knowledge graph to obtain an updated set of entity nodes and corresponding relationship edges; identifying temporal conflicts based on the extracted temporal attribute information and updating the temporal attribute information to obtain a set of relationship edges with temporal consistency; persisting the updated set of entity nodes and the set of relationship edges with temporal consistency to a graph database through batch operations; and responding to query requests by performing temporally-aware retrieval in the dynamic knowledge graph based on the temporal attribute information of the relationship edges.

Techniques are disclosed for denoising videos. In some embodiments, video frames are denoised using a denoising model that includes an encoder-decoder architecture and attention modules. During training of the denoising model, the attention modules learn weightings to upweight certain dimensions of input features to help pixel registration, remove ghosting artifacts, and improve temporal consistency when the frames of a video are being denoised. The denoising model can also be used to train a student denoising model that has a same architecture as, but is smaller and faster than, the denoising model. After training, noisy video frames can be input into the denoising model and / or the student denoising model to generate corresponding denoised video frames.

This invention discloses a low-latency, low-communication-overhead navigation method for resisting sudden link interruptions, relating to the fields of wireless communication, cooperative positioning, and graph signal processing. The invention constructs a momentum-accelerated graph filter in a distributed network, accelerating convergence by introducing an inertial momentum term into node state updates, and adaptively adjusting the polynomial truncation order using residual energy detection to reduce latency and communication overhead. Simultaneously, a spatiotemporal two-dimensional trust region model is constructed, combining temporal consistency and spatial geometric verification to perform real-time trust scoring of neighboring node links. Finally, based on the trust score, a robust M-estimator based on the Huber kernel function is used to update the graph displacement operator. This invention solves the problems of slow convergence and high communication overhead in traditional algorithms, and can effectively cope with sudden link interruptions and measurement drift in highly dynamic scenarios, achieving high-precision, highly robust distributed cooperative navigation.

The present application relates to the technical field of bridge monitoring, and more particularly to a multi-factor abnormal grading alarm method for bridge health state, which comprises: collecting response time series data through a bridge monitoring sensor to obtain basic monitoring data of the bridge; obtaining a first optimization factor by dynamically matching and reconstructing the response waveform of the adjacent sensor; obtaining a second optimization factor by measuring the spatio-temporal consistency of the transmission parameter difference of the upstream and downstream sensors; obtaining a dynamic equivalent damage value by jointly weighting the first and second optimization factors; and obtaining a multi-factor abnormal grading alarm result by multi-source fusion analysis of the dynamic equivalent damage value, thereby solving the problem that the prior art cannot identify the time series transmission rule distortion under a dynamic load scenario, resulting in early damage missed reports.

An image processing device comprises a memory and at least one processor, wherein the at least one processor is configured to: acquire a low-resolution video; acquire a plurality of first vector blocks through temporal division and spatial division of a low-resolution vector converted from the low-resolution video; input the plurality of first vector blocks into a denoising neural network, thereby acquiring a plurality of second vector blocks respectively corresponding to the plurality of first vector blocks and used to generate a high-resolution video; and output a high-resolution video generated by converting the plurality of second vector blocks, wherein, in a process of sequentially processing the plurality of first vector blocks, the denoising neural network acquires the plurality of second vector blocks on the basis of temporal consistency information obtained using an adjacent first vector block and spatial consistency information obtained using the low-resolution vector.

This invention provides a method for recognizing the operating status of solar thermal collectors based on a structure-aware segmentation network, belonging to the field of solar thermal equipment detection technology based on computer vision. First, it acquires equipment images and video data under different operating conditions through multi-source visual acquisition methods, and constructs a multidimensional dataset containing structural semantic labels, status labels, and temporal labels. Then, it builds a structure-aware semantic segmentation model based on an improved DeepLab network, and achieves high-precision segmentation of key structural regions of the thermal collector by introducing structural prior coding and structural constraint attention mechanisms. On this basis, it combines a temporal consistency discrimination module to model structural changes in continuous video frames, achieving dynamic recognition of states such as occlusion, offset, damage, and abnormal installation angles. Finally, it fuses the segmentation results and status discrimination results to output the equipment operating status and anomaly type, realizing automated detection and analysis.

Systems and methods presenting improvements in encoding and decoding of video for machine-based applications are provided which include modified feature conversion with non-linear transform of tensors of features and / or temporal consistency processing of tensors at an encoder and complimentary inverse processes at the decoder.

A system for processing and integrating contextual information across multiple time scales simultaneously analyzes immediate and historical data. The system comprises three main components working together to maintain temporal coherence. A temporal processing engine processes input data at different time scales using multiple processors, each dedicated to a specific time horizon. A context integration engine extracts and organizes contextual features from the input data, generating representations for different types of context while maintaining state information across these context types. A fusion engine combines the outputs from the temporal processors and integrates them with the context representations to generate coherent, unified outputs. The three engines operate cooperatively to ensure temporal consistency while processing context across multiple time scales. This architecture enables robust handling of complex temporal and contextual relationships, supporting applications in fields such as customer service, healthcare monitoring, and interactive systems where maintaining coherence across time scales is crucial.

This invention relates to a lightweight 4D occupancy prediction method for autonomous driving, belonging to the field of autonomous driving technology. The method first constructs a dynamic region identification and feature enhancement module, generating a dynamic mask using historical occupancy changes and motion intensity. A flow-guided temporal refinement mechanism is introduced to perform temporal alignment and residual optimization on predicted features at adjacent times, thereby improving the temporal consistency and dynamic target representation capability of future occupancy predictions. A perception constraint and joint optimization method oriented towards safety-critical areas is employed. A spatial distance weighting mechanism based on the driving risk potential field is constructed to strengthen near-field high-risk area supervision, and a voxel-level adaptive weighting is designed in conjunction with semantic distribution priors. Simultaneously, a multi-step consistency constraint based on temporal flow alignment is introduced to improve the prediction stability and reliability of critical areas and important traffic participants. Finally, the model is exported to achieve efficient inference, outputting voxel occupancy prediction results for the current and future times.

An electronic control unit that identifies multiple workloads from the execution of a runtime application, with each identified workload associated with a set of cache rows from a cache resource. A priority assignment for the set of cache rows is determined, at least in part, based on the expected usage of that set of cache rows during an upcoming time interval. In examples, the usage is based, at least in part, on memory access operations expected to be performed by the multitude of workloads during the upcoming time interval.

This invention discloses a spatiotemporal reconstruction method for three-dimensional ocean wave fields based on dispersion propagation and multi-frame fusion. Belonging to the field of marine observation and three-dimensional reconstruction technology, its operational steps include: sparse completion and dispersion propagation, weighted fusion and coarse physical estimation, residual refinement, and observation closed-loop dispersion consistency constraints. This invention features observation closed-loop dispersion consistency constraints (providing non-trivial physical consistency supervision signals, forcing the network to learn wave field representations that conform to physical laws, improving the temporal consistency of reconstruction results), residual learning and skip connections (solving the signal vanishing problem in deep sparse convolutional networks, enabling the refinement network to function effectively), joint optimization of multiple physical constraints (five loss functions work synergistically to constrain reconstruction quality from multiple dimensions), a supervised training strategy based on random degradation (eliminating the need for additional manual point-by-point annotation, reducing data acquisition costs), and automatic wave direction estimation (improving the automation and practicality of the system).

This invention relates to the field of video editing technology and discloses a method and system for enhancing temporal consistency in video editing. The method includes: processing the original video frame sequence in the original video stream data using an image diffusion model to obtain a set of frame-by-frame editing results; processing the set of frame-by-frame editing results using both the image diffusion model and the video diffusion model to obtain the frame-by-frame latent variables output by the image diffusion model and the video latent variables output by the video diffusion model, and performing feature processing to obtain initial noise for denoising; inputting the initial noise into the video diffusion model to obtain approximate video frames; calculating the optical flow field between adjacent frames of the approximate video frames using a pre-trained optical flow network; determining the fusion latent variables based on the optical flow field; and decoding the fusion latent variables to obtain the final edited video frame sequence. This invention solves the problems of temporal instability and low processing efficiency in existing temporal consistency enhancement methods by using a multi-agent collaborative mechanism to perform closed-loop organization and iterative optimization of the generation, evaluation, and parameter adjustment processes.

Human 3D pose can be estimated through triangulation of 2D representations of key points (e.g., joint positions). Triangulation can utilize various metrics, such as confidence scores associated with the 2D representations of key points and / or temporal consistency among multiple 3D representations of key points. 2D representations can be grouped, and candidate 3D representations for each group can be determined by considering the confidence scores of individual 2D representations within the group. The candidate 3D representation with the minimum error can then be used to characterize the key points. Further 3D representations of key points determined from images taken at different times can be used to refine the 3D representation of the key points.

An electronic control unit in which a plurality of workloads are identified from execution of a runtime application, where each identified workload is associated with a set of cache lines of a cache resource. A priority designation of the set of cache lines is determined based at least in part on a usage of the set of cache lines during an upcoming time interval. In examples, the usage is based at least in part on memory access operations that are expected to be performed by the plurality of workloads during the upcoming time interval.

This invention proposes a method, device, and readable storage medium for testing the time consistency of digital therapy applications. Addressing the issues of the inability to simulate the time effects of real devices and the lack of clinical constraint verification in long-term digital therapy testing, this method parses configuration information to obtain a set of time attributes for treatment tasks; constructs a virtual timeline engine to run an independent simulated clock, using acceleration factors and superimposed clock drift to simulate time base deviation; employs system API hooks and middleware proxies to construct a transparent proxy layer, intercepting and redirecting the clock interface to achieve decoupled injection of the simulated clock and the absolute reference clock; and synchronously collects application-aware timestamps and real timestamps of trigger events during operation, automatically verifying time consistency based on clinical time-series constraint rules. This invention is primarily used for reliability verification and compliance certification of time-sensitive medical digital therapy software.

This application discloses a method for cross-module reasoning result consistency verification in an AI-based centralized procurement governance hub. The method includes: acquiring reasoning results generated by AI models of multiple business modules within the centralized procurement governance hub based on procurement governance tasks; performing logical consistency verification on multiple reasoning results based on preset logical consistency verification rules or constraints, and performing temporal consistency verification on multiple reasoning results based on the generation timestamps of each reasoning result and the execution dependency order or path relationship between each business module; triggering a conflict handling mechanism when conflicts or inconsistencies are detected, and performing conflict rollback and result correction processing on conflicting reasoning results to obtain a verified and corrected consistent governance result; and integrating the various reasoning results and the consistent governance result to obtain a procurement governance decision. This method can improve the reliability, auditability, and traceability of procurement governance decisions.

This invention discloses a dynamic 3D target detection method and device based on spatiotemporal consistency multimodal information fusion, comprising two stages: model construction and inference. The model construction employs a four-stage strategy: independent pre-training of the point cloud branch with geometric consistency constraints; independent pre-training of the image branch with modal reliability constraints; multimodal collaborative training with cross-modal consistency constraints; and multi-frame temporal training with the encoder frozen, training only the dual-attention fusion module with temporal consistency constraints. In the inference stage, the trained model is used to: acquire current and historical frame data, extract point cloud BEV features and image BEV features, fuse them, perform spatiotemporal alignment, and then perform temporal fusion through the dual-attention fusion module to generate fused BEV features, which are then input into the target detection network to output the target category and 3D position. This invention effectively suppresses the accumulation of temporal errors and improves detection accuracy.

The application provides an event-triggered multi-agent predefined time consistency control method, relates to the technical field of multi-agent system cooperative control, and comprises the following steps: constructing the communication topology of a multi-agent system based on algebraic graph theory; establishing a predefined time convergence theorem as the core of realizing the predefined time consistency control of the multi-agent system; constructing a dynamic model of the multi-agent system to provide a basis for control algorithm design; designing an agent predefined time event-triggered control law to reduce the occupation of communication resources and realize the predefined time consistency control; and strictly proving the stability and predefined time convergence of the multi-agent system. Through the predefined time convergence theorem, the upper limit of the convergence time is set as an adjustable parameter, so that the state of the agent converges to the small neighborhood of the average value of the initial state within the predefined time, and meanwhile, the event-triggered control mechanism is combined to effectively reduce the communication frequency between the agents and improve the overall operation and communication transmission efficiency of the system.

A system and a method are disclosed. The method includes extracting a plurality of video frames including at least a first frame and a second frame; generating a predicted soft volume for the second frame; generating a weight map of pixels for the first frame; combining the predicted soft volume for the second frame with the weight map of pixels for the first frame to generate combined predictions for the output frame; and generating a video including the combined predictions for the output frame.